Joel Jayme T. Jayme, Godwin S. Monserate
This study investigates the feasibility of utilizing YOLOv8 (You Only Look Once version 8) for facial recognition with novel datasets. After the global pandemic, there is a pressing need for alternative biometric technologies that minimize physical contact and address hygiene concerns. Traditional fingerprint biometrics carry risks of disease transmission and security vulnerabilities, prompting the exploration of other biometric authentication methods. YOLOv8, known for its real-time object detection capabilities, is evaluated for its high performance in facial detection and recognition tasks. The research examines YOLOv8's accuracy, precision, and reliability, Also the Researcher will use Stack ensemble and its potential benefits in helping the base Model Detector (YoloV8) with a relatively small dataset and classes preventing it from Generalization as well as its potential benefits in enhancing hygiene and security compared to traditional fingerprint biometrics. The researcher uses a novel dataset using approximately 1000 images in order to test whether or not the YoloV8 with ensemble is capable or better at performing by using this metric mean average precision(mAP) 91.4%, Recall 82.6%, Accuracy of 94.1% as a baseline for its performance the researcher can reliably compare the result with scientific and mathematical backings. Ultimately, this research contributes to the advancement of biometric authentication technology, offering potential solutions to concerns brought about by the pandemic and the evolving landscape of cybersecurity threats. © 2025 IEEE.
University of San Carlos, Department of Computer, Information Sciences and Mathematics, Cebu, Philippines
